The states of water in Norway spruce (Picea abies (L.) Karst.) studied by low-field nuclear magnetic resonance (LFNMR) relaxometry: assignment of free-water populations based on quantitative wood anatomy

Introduction

The moisture content (MC) as well as the location of the water, i.e. bound water in the cell wall or free water in voids such as cell lumina or bordered pits, influence wood degradation by wood-degrading fungi (Ibach 2013) as well as most physical wood properties such as strength and dimensional stability (Shmulsky and Jones 2011). Quantitative and qualitative methods for moisture characterization are consequently of high value.

Low-field nuclear magnetic resonance (LFNMR) has previously been used to study water in wood, both to determine MC (Sharp et al. 1978; Hartley et al. 1994, 1996; Merela et al. 2009) and to differentiate between different water states or populations (Menon et al. 1987, 1989; Flibotte et al. 1990; Araujo et al. 1992, 1994; Labbé et al. 2002, 2006; Almeida et al. 2007; Thygesen and Elder 2008, 2009). The LFNMR approach is decribed by Rutledge (1992). In short, the spin-spin relaxation time (T₂) determined by LFNMR is a measure of how tight the water is bound: the more tight the bonding, the shorter the T₂. Provided that the pore wall surfaces are similar throughout the material, the T₂ of the free water is related to the surface area to pore volume ratio where liquid water in smaller pores have a shorter T₂ than in larger pores (Brownstein and Tarr 1979; Menon et al. 1987; Almeida et al. 2007). The T₂ of free water is not only related to the pore-radius and pore-wall properties, but can also be influenced by the presence of solutes (Hsieh et al. 2014).

In LFNMR studies on softwoods, three or four peaks representing different water populations have been identified in water-saturated or green specimens (Menon et al. 1987, 1989; Flibotte et al. 1990; Araujo et al. 1992; Labbé et al. 2002, 2006; Thygesen and Elder 2008). The 1^st peak, generally at <3 ms (Menon et al. 1987; Araujo et al. 1992, 1994; Labbé et al. 2002, 2006; Thygesen and Elder 2008; Telkki et al. 2013) is assigned to the water bound in the cell wall, while peaks with a longer T₂ are assigned to free water. This differentiation between bound water and free water is made because the former has a T₂ which is constant and independent of MC changes above the fibre saturation point (FSP) (Araujo et al. 1992). Below the FSP, however, the amplitude (Araujo et al. 1992) as well as the T₂ of this peak decreases with decreasing MC (Menon et al. 1987; Araujo et al. 1994; Almeida et al. 2007). Peaks with longer T₂ are attributed to free water, because their amplitudes decrease with decreasing MC above the FSP (Araujo et al. 1992). In addition, a 4^th peak with a longer T₂ was detected in two studies (Labbé et al. 2006; Thygesen and Elder 2008). Labbé et al. (2006) attributed this peak to organic compounds.

In the studies where two peaks were found for free water, these were in the ranges 9–80 ms and 30–400 ms, respectively (Menon et al. 1987; Flibotte et al. 1990; Araujo et al. 1992; Labbé et al. 2002, 2006). As the pore size influences the T₂, the free-water peak with the shortest T₂ (9–80 ms) has been attributed to water in latewood (LW) lumen and ray cells, while the free-water peak with the longer T₂ (30–400 ms) has been attributed to water in the larger earlywood (EW) lumina (Menon et al. 1987, 1989). A shorter T₂ in LW than in EW was also observed by magnetic resonance imaging (MRI) (Javed Muhammad et al. 2015).

So far, the assignment of free-water populations within wood has been made by qualitatively sorting and pairing the observed T₂ values with the void sizes present within the specimens because water inside smaller voids generally has a shorter T₂. However, in the present study, the assignment is based on relative amplitudes/peak areas obtained from LFNMR data of Norway spruce (Piceaabies (L.) Karst.) and the void volumes available for free water. That is, rather than relying on observed T₂ values only, the information captured by the LFNMR measurement regarding the amount of water in each of these populations was also taken into account. In addition, to reduce the number of possible void types, well-defined specimens were investigated. Furthermore, the specimens were measured both before and after extraction to identify possible effects of extractives. Finally, LFNMR data were compared with sorption isotherms determined for all wood types to study possible differences in MCs within the cell wall.

Materials and methods

Materials: Samples were taken from one Norway spruce (Picea abies(L.) Karst) tree from a stand in the southern part of Sweden, see Fredriksson et al. (2016) for details. Specimens were cut from a total of 30 individual growth rings of mature sapwood (sW), mature heartwood (hW) and juvenile hW, i.e. 10 growth rings of each wood type. For each wood type, specimens were taken from two different boards. EW and LW were separated; hence, a total of 60 specimens were prepared (Table 1).

All specimens were cut to 10 mm in the longitudinal direction. The cross section of each specimen was cut to fit into the NMR tube (inner diameter 8 mm) so that the vacuum-saturated mass of each specimen was about 0.15 g. Because of the thin LW rings, two LW pieces were needed to obtain approximately the same mass as one EW specimen. All specimens were vacuum-saturated in vacuum (0–1 mbar) for 1 h. Deionized water was then let in and finally atmospheric pressure was applied.

NMR: The measurements were performed on vacuum-saturated specimens. Each specimen was wiped on a wet piece of paper to avoid the presence of free water in the NMR tube, but without drying the specimen. In the case of LW, where two pieces were needed, both pieces were wiped individually to avoid free water between them. The specimens were then placed in the NMR tubes, which were sealed with a cap and placed in the NMR instrument (mq20-Minispec, Bruker, Billerica, MA, USA) with a 0.47-Tesla permanent magnet, which corresponds to a 20-MHz proton resonance frequency. Care was taken so that the specimen orientation was the same for all measurements. The spin-spin relaxation time (T₂) was determined using the Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence with a pulse separation (τ) of 0.1 ms, 8000 echoes, gain 82 dB, 32 scans and a recycle delay of 5 s. The 90° and 180° radio frequency pulses in the instrument have the same radio frequency gain and hence the 180° is about twice the duration of the 90° pulse (3.54 μs and 7.22 μs, respectively). The magnet temperature was 40°C, but the probe used enabled temperature control of the sample tube by water circulation. The sample temperature was set to 20°C. As the room temperature was 23°C, the specimen was kept in the NMR for 5 min before the measurement started. The mass of each specimen was determined before each NMR measurement.

After the measurements, the specimens were dried, first above water in a desiccator (i.e. close to 100% relative humidity, RH) for 4 weeks and then at 45°C for 19 days. Hereafter, the equilibrium mass at 45°C was determined. Half of the specimens were then extracted in ethanol for 24 h in a Soxhlet apparatus. These specimens were then again dried at 45°C for 8 weeks and their mass was determined. The extracted specimens were then vacuum-saturated by the procedure described above, and a second set of NMR measurements was performed. Finally, all specimens were dried at 105°C for 24 h, the dry mass (m_dry) was determined and the MC (u) was calculated as:

where m_eq is the mass of the vacuum-saturated specimens. For the extracted specimens, the dry mass was adjusted in order to calculate the MC on the same terms as for the non-extracted specimens. The mass of the extractives (m_extr) was determined as:

where m_{1, 45°C} is the equilibrium mass at 45°C before extraction and m_{2, 45°C} is the equilibrium mass at 45°C after extraction. The dry mass of the extracted samples was determined by adding the mass of the extractives m_extr to the dry mass of the extracted specimen determined at 105°C. The average extractive content in relation to the dry mass was about 2%, which is in agreement with literature data for P. abies (Fengel and Wegener 2003). The average MC after vacuum saturation for each group of specimens is given in Table 1.

Data evaluation: The CPMG data were normalized by dividing the relative intensity by the vacuum-saturated mass of each specimen. These data were then evaluated using two methods: discrete multi-exponential fitting (Pedersen et al. 2002) and the Laplacian transformation method Contin as described by Provencher (1982). For the discrete multi-exponential fitting the CPMG data were fitted to:

based on up to five components. Four components were chosen because there was only a minor decrease in the residual sum of squares when an additional component was added. The Contin method gives a continuous distribution of the relaxation times (T₂) for each specimen, 256 data points were used in the 0.01–10 000 ms range. From this continuous distribution, the T₂ value corresponding to the maximum amplitude of each peak was determined. Here, five peaks were identified, but all peaks were not present for all specimens. The T₂ values for the different peaks for both the multi-exponential and the Contin fit were analysed by analysis of variance (ANOVA) to establish differences between the different groups of specimens (EW/LW, mature/juvenile wood, hW/sW) and the influence of extraction. For ANOVA, the Statistics Toolbox in MATLAB (MATLAB R2013b, The MathWorks, Inc., Natick, MA, USA) was used. In addition, the relative amplitudes (multi-exponential fit) and the relative area under each peak (Contin) were evaluated. The latter was evaluated as dS/d(lnT₂) (Terenzi et al. 2015).

Volume estimation of spaces in the wood structure available for free water: The amplitude of each T₂ component represents the relative amount of water (McDonald et al. 2010) and for a continuous curve-fit, the area under the curve is proportional to the amount of water (Labbé et al. 2006). To clarify which water the two peaks related to free water represent, the volumes of all spaces in the wood structure that contain free water were determined. Dimensions of the lumen cross section of mature EW and LW were estimated based on images taken with a scanning electron microscope (SEM, SU3500, Hitachi High-Technologies Corporation, Krefeld, Germany). The specimens were prepared by cutting thin slices with a razor blade, mounting them on sample holders and finally sputter coating the samples with gold (Cressington 108 Auto Sputter Coater, Cressington Scientific Instruments Ltd., Watford, UK). The other dimensions and parameters are based on literature data as presented in Table 2. The void volumes available for free water were calculated with the Eqs. (4–21); all parameters in these equations are defined in Figure 1.

Figure 1:

Definition of all parameters used in Eqs. (4–21) for determination of the void volumes in the wood structure available for free water.

The volume of the tracheid lumen, V_lumen, was determined by assuming a rectangular cross section:

As the tracheid ends are tapered, the volume of the tracheid ends was determined separately as:

The volume of the pit chamber was assumed to be the same irrespective of whether the pit was aspirated or not. The volume was determined as twice the volume of a spherical cap:

As the cell walls of LW are thicker, the aperture of the bordered pits resembles a capillary rather than an orifice (Siau 1984); it is thus possible that water is held here. The aperture was approximated to be cylindrical and the volume was calculated:

Norway spruce contains both parenchymatous and tracheidal ray cells (Back 1958). Both the length (Back 1958) and the types of pits (Nyrén and Back 1960) differ between these two cell types, and they were therefore considered separately in the volume calculations. The lumen of the ray cells was determined by assuming a rectangular cross section:

where l_ray is the length of the tracheidal (l_{ray, rt}) and parenchymatous (l_{ray, rp}) ray cells, respectively. The volume of the simple pits, V_{pit, rp}, connecting the parenchymatous ray cells was determined by:

The volume of the pit chamber of the bordered pits connecting the tracheidal ray cells, V_{pit, rt}, was determined as for the bordered pits in the tracheids by Eq. (6). In the literature, relative volumes of tracheid cells and ray cells can be found. However, as the numbers of bordered pits and tracheid ends depend on the number of cells, the numbers of tracheid cells and ray cells were estimated. The volume of one ray cell (including the cell wall) was estimated by:

The number of ray cells was then estimated by:

where p_r is the relative volume of ray cells. The volume of one EW and LW cell, respectively, was estimated by:

The number of tracheids in the EW and LW, respectively, could then be estimated as:

The total volumes of the tracheid lumen V_{tot, lumen,} the tracheid lumen ends V_{tot, ends}, the bordered pits V_{tot, pit}, the apertures of the bordered pits in the LW V_{tot, ap}, the tracheidal ray cell lumina V_{tot, lumen, rt}, the parenchymatuous ray cell lumina V_{tot, lumen, rp}, the pits in the tracheidal ray cells V_{tot, pit, rt} and the pits in the parenchymatous ray cells V_{tot, pit, rp} were finally estimated for EW and LW, respectively, as:

The relative volumes were estimated by dividing the volume of each space by the total volume, i.e. the sum of Eqs. (14–21).

Initially, all volumes were estimated based on the mean values given in Table 2. However, the width and the length of the tracheid and the ray cells as well as the size of h_end were also varied between the minimum and maximum (Table 2) to get an idea of how much the individual dimensions influence the results. The parameters were varied one at a time.

Sorption isotherms: Measurements were performed in a sorption balance (dynamic vapour sorption, DVS Advantage, Surface Measurement Systems Ltd., London, UK). Here, the RH is regulated by changing the proportions of dry and water-saturated nitrogen gas. The sorption balance monitors the mass of the specimen with a resolution of 0.1 μg while the relative humidity is incrementally changed. The method is further described by Williams (1995). Small specimens were cut from the vacuum-saturated specimens prepared for the NMR measurements. For water removal from the specimen surface, the specimen was wiped by a moist cloth before it was placed in the sorption balance. The specimens were equilibrated to the following RH-levels: 97, 95, 90, 85, 80, 70, 60, 50, 40, 20, 10, 5 and 0%. The RH was then increased again in the same steps. The measurements were performed at 20°C and the equilibrium criterion (dm/dt) was set to 0.001% min⁻¹ over a 10-min period. The dry mass of each specimen was 4–10 mg.

Results and discussion

Volumes in the wood structure available for free water

The void volumes in the wood structure available for free water estimated based on the mean values in Table 2 are shown in Table 3. In addition, one parameter at a time was varied between the minimum and maximum values (Table 2) to understand the influence of individual parameters. In Figure 2 it is seen that the tracheid length and width had the largest influence on the relative volumes. For LW, also the ray cell dimensions had an influence.

Table 3:

Relative void volumes available for free water within mature wood and juvenile wood, respectively.

Cell and void type	Mature wood		Juvenile wood
	EW	LW	EW	LW
Tracheids
Lumen (%)	89.72	82.00	84.73	78.64
Lumen ends (%)	2.68	1.94	4.74	3.45
Bordered pits (%)	2.09	1.68	4.91	2.77
Aperture, LW pits (%)		0.07		0.22
Ray cells
Lumen (parenchyma) (%)	2.67	6.95	2.73	7.25
Lumen (tracheids) (%)	1.57	4.08	1.60	4.26
Simple pits (%)	0.14	0.36	0.14	0.37
Bordered pits (%)	1.12	2.92	1.15	3.04

Figure 2:

The influence of the tracheid and ray cell length and width as well as the influence of the size of h_end on the volumes available for free water for mature earlywood (EW) (a), mature latewood (LW) (b), juvenile earlywood (EW) (c) and juvenile latewood (LW) (d).

The asterisk represents the relative volume determined by the mean values in Table 2. The influence of individual parameters on the volumes (squares, circles and triangles) was determined using the min and max values given in Table 2.

Sorption isotherms

The results of the sorption balance measurements (Figure 3) show that the sorption isotherms are similar for the different wood types. However, as reported in previous studies (Hill et al. 2011; Sharratt et al. 2011), small differences were seen at high RH-levels. In the present study, it is not possible to ascribe these small differences to actual differences of the moisture sorption properties of the wood. Instead, it is likely that these differences are due to methodical aspects. As in most other DVS-studies, the dm/dt-criterion was the basis for equilibrium determination at each RH-level. As dm (%) is determined with the initial mass as a reference, the dm/dt-criterion does not work equally well at all RH-levels. When plotting mass as a function of time it was seen that the sample is not perfectly equilibrated at higher RH-levels.

Figure 3:

The sorption isotherms for earlywood (EW) and latewood (LW) of mature sapwood (sW) (a) and mature (b) and juvenile heartwood (hW) (c). All data from a, b and c are shown together in d, for legend see a–c.

Note that two replicates were used for each wood type, which is not seen in the figures since the spread was very small.

LFNMR studies

The continuous T₂ distributions are shown in Figure 4, where the filled grey range represents the mean values with the StDs and the bars represent the components of the discrete multi-exponential fit. The relative amplitudes of the latter and the relative areas under the different peaks according to the Contin method are listed in Table 4, while the corresponding T₂ values for each peak are shown in Table 5. In general, both curve-fit methods gave the same results in terms of T₂ for the different peaks; the maxima of the continuous distribution coincided with the mean T₂ for the discrete components for Peaks 1 and 3. However, for Peaks 2 and 4 deviations were seen; the T₂ for Peaks 2 and Peak 4 were generally shorter and longer for the continuous distribution, respectively (Figure 4, Table 5).

Figure 4:

The continuous T₂ distribution and the fit to four discrete components (bars) are shown for earlywood (EW) (a, c, e) and latewood (LW) (b, d, f) of mature sapwood (sW) and mature and juvenile heartwood (hW).

The width of the bars and the filled area of the continuous distribution represent the mean value including the Std. The error bars on the discrete components show the Std of each amplitude. The error bar is placed at the mean T₂. The numbers in a and b refer to the numbering of the peaks used in the present study.

Table 4:

The mean relative amplitudes (discrete multi-exponential fit) and the relative areas under each peak (continuous fit).

	Peak 0		Peak 1	Peak 2		Peak 3	Peak 4
Discrete multi-exponenital fit
Non-extracted
Mature sW
EW			0.10	0.06		0.65	0.19
LW			0.25	0.16		0.50	0.10
Mature hW
EW			0.10	0.08		0.75	0.07
LW			0.26	0.16		0.48	0.11
Juvenile hW
EW			0.10	0.16		0.65	0.09
LW			0.20	0.25		0.46	0.09
Extracted
Mature sW
EW			0.10	0.09		0.74	0.07
LW			0.28	0.18		0.44	0.10
Mature hW
EW			0.10	0.09		0.75	0.06
LW			0.29	0.17		0.47	0.08
Juvenile hW
EW			0.10	0.12		0.75	0.03
LW			0.22	0.22		0.50	0.06
Continuous fit
Non-extracted
Mature sW
EW	0.00	(3/10)	0.10	0.04	(10/10)	0.85	0.02	(8/10)
LW	0.02	(1/10)	0.25	0.02	(7/10)	0.65	0.06	(10/10)
Mature hW
EW	0.03	(3/10)	0.09	0.03	(10/10)	0.83	0.02	(10/10)
LW	0.00	(0/10)	0.27	0.01	(6/10)	0.63	0.09	(10/10)
Juvenile hW
EW	0.00	(0/10)	0.10	0.02	(9/10)	0.85	0.03	(9/10)
LW	0.00	(1/10)	0.21	0.01	(2/10)	0.70	0.08	(10/10)
Extracted
Mature sW
EW	0.01	(2/5)	0.09	0.04	(5/5)	0.83	0.02	(5/5)
LW	0.00	(0/5)	0.29	0.03	(3/5)	0.59	0.10	(5/5)
Mature hW
EW	0.00	(2/5)	0.09	0.03	(5/5)	0.85	0.02	(5/5)
LW	0.00	(0/5)	0.31	0.00	(1/5)	0.64	0.06	(5/5)
Juvenile hW
EW	0.00	(0/5)	0.09	0.03	(5/5)	0.86	0.01	(5/5)
LW	0.00	(0/5)	0.25	0.00	(1/5)	0.72	0.03	(5/5)

The relative amplitudes/areas were determined as the amplitude/area for each peak divided by the total amplitude/area. For the continuous curve-fit, Peaks 0, 2 and 4 did not occur for all specimens. For these peaks, the data in brackets mean the number of specimens for which this peak occurred in relation to the total number of specimens.

Table 5:

Mean values of the T₂ (ms) determined by fitting to four discrete components and by a continuous curve-fit.

	Peak 0			Peak 1		Peak 2			Peak 3		Peak 4
	mean	std		mean	std	mean	std		mean	std	mean	std
Discrete multi-exponenital fit
Non-extracted
Mature sW
EW				1.7	0.3	19.6	8.4		85.2	10.3	226.7	102.9
LW				1.9	0.1	25.1	5.2		74.4	9.9	357.7	162.7
Mature hW
EW				1.8	0.1	29.6	7.7		95.4	5.6	350.0	92.7
LW				1.9	0.3	29.7	5.2		78.1	10.5	438.6	116.5
Juvenile hW
EW				1.6	0.2	31.4	7.6		72.0	9.5	436.6	133.8
LW				1.5	0.1	26.3	1.8		67.5	6.7	391.7	117.7
Extracted
Mature sW
EW				2.0	0.4	27.3	10.1		91.9	8.9	374.3	137.0
LW				1.8	0.1	24.8	3.0		71.5	10.3	508.5	33.4
Mature hW
EW				2.0	0.3	29.9	6.3		93.5	5.6	350.8	95.8
LW				1.9	0.2	26.1	2.7		69.6	8.5	346.1	150.1
Juvenile hW
EW				1.8	0.0	26.4	4.5		72.5	5.7	307.9	55.7
LW				1.7	0.4	21.8	2.5		60.2	3.9	278.4	82.0
Continious fit
Non-extracted
Mature sW
EW	0.3	0.1	(3/10)	1.5	0.1	9.7	2.2	(10/10)	94.1	9.7	1143.5	1279.9	(8/10)
LW	0.1	0.0	(1/10)	1.7	0.1	13.6	4.4	(7/10)	67.1	8.0	1000.9	1696.7	(10/10)
Mature hW
EW	0.2	0.1	(3/10)	1.5	0.2	8.3	2.1	(10/10)	93.5	7.9	730.8	202.5	(10/10)
LW	–	–	(0/10)	1.9	0.3	13.8	4.4	(6/10)	67.8	9.6	491.5	150.8	(10/10)
Juvenile hW
EW	–	–	(0/10)	1.3	0.2	8.1	2.0	(9/10)	66.5	8.9	599.5	86.3	(9/10)
LW	0.0	0.0	(1/10)	1.2	0.2	7.7	2.9	(2/10)	51.4	6.0	443.7	120.3	(10/10)
Extracted
Mature sW
EW	0.2	0.2	(2/5)	1.5	0.2	9.4	1.6	(5/5)	90.8	4.1	727.5	215.4	(5/5)
LW	–	–	(0/5)	1.7	0.1	14.2	4.9	(3/5)	60.9	10.8	545.2	65.1	(5/5)
Mature hW
EW	0.1	0.0	(2/5)	1.6	0.3	9.2	0.9	(5/5)	90.9	7.2	1469.4	1665.2	(5/5)
LW	–	–	(0/5)	1.8	0.2	9.7	0.0	(1/5)	59.6	5.6	430.5	184.8	(5/5)
Juvenile hW
EW	–	–	(0/5)	1.4	0.1	8.0	2.1	(5/5)	68.8	7.7	719.8	273.8	(5/5)
LW	0.1	0.0	(0/5)	1.7	0.4	9.7	0.0	(1/5)	48.8	7.1	402.5	192.2	(5/5)

For the continuous curve-fit, Peaks 0, 2 and 4 did not occur for all specimens. For these peaks, the number of specimens for which this peak occurred in relation to the total number of specimens is given in brackets.

The P-values from the ANOVA are presented in Table 6. There were significant differences between EW and LW for Peaks 1, 2 and 3, and between mature and juvenile wood for Peaks 1 and 3. No significant differences were seen between hW and sW. Regarding the influence of extraction, the T₂ of Peak 1 was generally slightly longer and the T₂ of Peak 3 was generally slightly shorter for the extracted specimens. These differences were, however, only significant for one of the curve-fit methods (Table 6). Otherwise, the same significant differences between groups were generally seen, irrespective of the curve-fit method (continuous distribution or the discrete components).

Table 6:

The P-values from ANOVA calculated for T₂ values from the continuous curve-fit (cont) and the discrete multi-exponential fit (exp), respectively.

Factors	Curve-fit	ANOVA P-values
		Peak 1	Peak 2	Peak 3	Peak 4
hW/sW	Cont	0.333	0.254	0.947	0.370
	Exp	0.408	0.684	0.635	1.000
Mature/juvenile wood	Cont	0.000a	0.124	0.000a	0.378
	Exp	0.000a	0.827	0.000a	0.225
EW/LW	Cont	0.000a	0.000a	0.000a	0.106
	Exp	0.009a	0.001a	0.000a	0.322
Non-extr./extr.	Cont	0.035a	0.998	0.071	0.914
	Exp	0.057	0.250	0.011a	0.483

The data with ^a designate significant differences (P<0.05).

Water populations of Peaks 2 and 3

The T₂ corresponding to Peak 3 (Table 5) is in the range that was previously assigned to lumen water in the tracheid cells (Menon et al. 1987; Araujo et al. 1992; Labbé et al. 2002, 2006; Thygesen and Elder 2008). In the present study, there was a significant difference in T₂ between EW, LW, juvenile and mature wood (Table 6). The T₂ of Peak 3 was shorter for LW, which is in agreement with previous studies (Menon et al. 1987, 1989; Javed Muhammad et al. 2015) and is due to the fact that LW lumina are smaller than those of EW. The T₂ of Peak 3 was also shorter for juvenile than for mature wood (Table 5), which is in agreement with results of Araujo et al. (1992). This is probably due to the smaller diameters in juvenile tracheids (Brändström 2001; Schweingruber 2007).

Peak 2 has previously been assigned to water in LW lumen and rays (Menon et al. 1987, 1989), but in the present study, where LW and EW were observed separately, this peak was seen for both the EW and the LW specimens. However, Peak 2 was not always seen for the LW specimens using the continuous curve-fit (Table 5), probably because it merged with Peak 3, which could happen as the T₂ of Peak 2 was longer and the T₂ of Peak 3 was shorter in LW.

To determine the fraction of each of the two free-water peaks, the amplitudes of Peaks 2 and 3 respectively from the discrete exponential fit was divided by the sum of the amplitudes of these peaks. Similarly, the area under Peaks 2 and 3 for the continuous curve-fit was divided by the sum of the areas of Peaks 2 and 3. The results represent the relative amount of free water represented by these peaks. In Figure 5, these relative amplitudes and areas representing Peaks 2 and 3 from the LFNMR measurements are compared with the volumes available for free water in the wood structure according to the estimates given in Table 3. Note that in Figure 5, the ray cell lumen volumes and the pit volumes were summed, i.e. the ray cell lumen is the total volume of the lumina for the two ray cell types and the pit volume is the sum of all volumes related to the pits in both the tracheid cells and the ray cells. For the discrete exponential fit for the EW, the relative amplitude of Peak 3 corresponded fairly well with the relative volume of the tracheid lumen (Figure 5). Thus, the relative amplitude of Peak 2 corresponds with the volume of smaller voids (ray cell lumen, pits and tracheid lumen ends) (Figure 5). However, for LW, especially for juvenile LW, the estimated lumen volume was larger than the relative amplitude of Peak 3. Here, a better correlation with the relative amplitude of Peak 3 would be obtained in case of decreasing tracheid width or the tracheid length (Figure 2). It is not obvious why the ray cell water should be included in Peak 2 for the LW; the length of the cells is different, but the width and their surface area to pore volume ratios are similar, and thus the T₂ values are also expected to be similar. One hypothesis is that organic substances in the ray cell lumina lower the T₂. If it is true, this effect should not be seen for the extracted specimens, but the results were similar also after extraction. Another possible explanation is that the ray cells are oriented perpendicular to the tracheid cells. For the cell wall water, the orientation of the wood sample during measurement has been shown to influence ¹H NMR data (Terenzi et al. 2013). However, irrespective of how the dimensions are varied between the minimum and maximum values in Table 2, it is not possible to assign the ray cell lumen water to Peak 3 (Figures 2 and 5).

Figure 5:

The volume bars represent the estimated volume fractions available for free water.

The relative amplitudes from the discrete exponential fit (exp) and the relative areas under the continuous curve-fit (cont) for the free-water peaks (Peaks 2 and 3) are also shown and show the relative amount of free water represented by Peaks 2 and 3, respectively. For mature wood, the values are average values for mature sW and hW. All LFNMR data are from non-extracted wood samples.

For the continuous curve-fit, the results were different; here the relative area under Peak 2 was small, especially for the LW, and only the water in the pits seemed to be included (Figure 5). For juvenile wood, the relative area for Peak 2 was even smaller than the volume fraction corresponding to the pits. For both EW and LW, the water in the ray cell lumen was included in Peak 3, i.e. water in the ray cell lumina does not seem to give a separate signal from water in the tracheid lumen even for the EW samples. Because of the small relative area under Peak 3, it was not possible to assign this peak differently irrespective of how the dimensions were varied (Table 2).

The results from the ANOVA showed that there was a significant difference between EW and LW in terms of the T₂ values for Peak 2 (Table 6). For the discrete multi-exponential fit, the overall mean T₂ was shorter for the LW specimens. This is logical as the void volumes (bordered pits, lumen), except for those present in ray cells, generally are smaller in LW and would thus give a shorter T₂. From Table 5, it is seen that this shorter T₂ for LW is most pronounced for the extracted specimens. However, for the continuous curve-fit, the T₂ for Peak 2 was generally shorter in case of EW. A comparison between the relative area of Peak 2 and the volume estimations indicates that Peak 2 represents the water in pits. This observation is counterintuitive because the pits are smaller in LW. However, for the continuous fit, Peak 2 was not present for all LW specimens and the mean values are thus based on fewer observations for LW. This is the reason why these results should be interpreted with care.

Correlating the T₂ value with the surface area to volume ratios of the voids would also offer a peak assignment. However, this approach requires that the surface relaxivity is the same throughout the material, which might not be the case due to differences in chemical composition in different parts of the wood structure.

Conclusions

There was a significant difference in T₂ of cell wall water between EW and LW as well as between juvenile and mature wood, but no differences were observable between the sorption isotherms of these anatomical parts of wood. Thus, the reasons for the T₂ differences are not clear, but different chemical cell wall composition or extractive contents are possible explanations. Two peaks corresponding to free water were seen by both the discrete and the continuous curve-fit for both EW and LW. However, when comparing LFNMR signal fractions to estimates of void volumes within the wood structure, the two peaks from the two curve-fit methods do not seem to represent the same water. The free-water peak with the shorter T₂ (Peak 2) constituted a smaller fraction of the total amount of free water for the continuous curve-fit than for the discrete multi-exponential fit, especially in case of LW specimens. For the discrete curve-fit, Peak 2 was assigned to water in the pits and the ray cell lumina, while the free-water peak with the longer T₂ (Peak 3) was assigned to water in the tracheid lumina. However, for the continuous curve-fit, Peak 2 seemed to include only the water in the pits, while the rest of the free water was included in Peak 3. A population with a longer relaxation time than lumen water was found to most likely correspond to water on the specimen’s surface.